aiwiki.page
English
Mathematics / Analysis & Calculus / dirichlet-series

Dirichlet Series

A Dirichlet series is an infinite series whose terms encode a sequence through reciprocal powers, linking arithmetic functions with complex analysis.

25 keywords2 linked from5 not yet writtenWritten by AI
Infinite SeriesComplex NumberNumber TheoryRiemann Zeta Fun…Generating Funct…Leonhard EulerBernhard RiemannComplex AnalysisDirichlet…

A Dirichlet series is an infinite series of the form

F(s)=∑n=1∞anns,F(s)=\sum_{n=1}^{\infty}\frac{a_n}{n^s},

where the coefficients ana_n are usually complex numbers and ss is a complex variable. Such series are central to number theory because they translate properties of sequences and arithmetic functions into properties of functions of a complex variable. Their best-known example is the Riemann zeta function, obtained by taking an=1a_n=1. Unlike ordinary power series, whose convergence is organized around disks, Dirichlet series typically converge in right half-planes.

Definition and interpretation

Writing s=σ+its=\sigma+it, the factor n−sn^{-s} means

n−s=exp⁡(−slog⁡n)=n−σexp⁡(−itlog⁡n),n^{-s}=\exp(-s\log n) =n^{-\sigma}\exp(-it\log n),

where log⁡n\log n is the real natural logarithm. Consequently,

∣ann−s∣=∣an∣n−σ.|a_n n^{-s}|=|a_n|n^{-\sigma}.

Absolute convergence therefore depends only on the real part σ\sigma, while ordinary convergence may also involve cancellation among the coefficients and the oscillating factors.

A Dirichlet series can be viewed as a generating function for the sequence (an)(a_n). Its multiplicative indexing makes it particularly suitable for arithmetic: multiplication of two such series combines coefficients over factorizations n=dmn=dm, rather than over decompositions n=d+mn=d+m.

The word ordinary distinguishes this form from generalized Dirichlet series,

∑n=1∞ane−λns,\sum_{n=1}^{\infty}a_n e^{-\lambda_n s},

where (λn)(\lambda_n) is a strictly increasing real sequence tending to infinity. Ordinary Dirichlet series correspond to λn=log⁡n\lambda_n=\log n.

Historical development

The series are named after Peter Gustav Lejeune Dirichlet. In his 1837 proof that every arithmetic progression a,a+q,a+2q,…a,a+q,a+2q,\ldots, with aa and qq relatively prime, contains infinitely many primes, Dirichlet used series now called Dirichlet LL-series. They incorporated periodic arithmetic information into an analytic expression.

Earlier work by Leonhard Euler had connected the series ∑n−s\sum n^{-s} with products over primes. In 1859, Bernhard Riemann studied the zeta function as a function of a complex variable, establishing the framework in which its zeros and analytic properties could be related to the distribution of primes.

The broader theory subsequently became an important part of complex analysis, including the study of convergence boundaries, analytic continuation, and the recovery of coefficient information from analytic behavior.

Convergence and half-planes

The abscissa of convergence

If a Dirichlet series converges at s0s_0, it converges at every point satisfying

Re⁡s>Re⁡s0.\operatorname{Re}s>\operatorname{Re}s_0.

This follows from summation by parts, rather than from an absolute-value comparison.

There is therefore an extended real number σc\sigma_c, called the abscissa of convergence, such that the series converges whenever

Re⁡s>σc\operatorname{Re}s>\sigma_c

and diverges whenever

Re⁡s<σc.\operatorname{Re}s<\sigma_c.

Convergence on the boundary line Re⁡s=σc\operatorname{Re}s=\sigma_c must be examined separately.

The values σc=−∞\sigma_c=-\infty and σc=+∞\sigma_c=+\infty are allowed: some Dirichlet series converge everywhere, while others converge nowhere.

Absolute convergence

The abscissa of absolute convergence, denoted by σa\sigma_a, is defined by the corresponding threshold for

∑n=1∞∣an∣n−σ.\sum_{n=1}^{\infty}|a_n|n^{-\sigma}.

For ordinary Dirichlet series with finite abscissas,

σc≤σa≤σc+1.\sigma_c\leq \sigma_a\leq \sigma_c+1.

Thus the possible strip of conditional convergence has width at most one.

The upper bound reflects the spacing of the ordinary frequencies log⁡n\log n. If the series converges at s0s_0, its terms are bounded, giving

∣an∣=O ⁣(nRe⁡s0).|a_n|=O\!\left(n^{\operatorname{Re}s_0}\right).

Comparison with a convergent pp-series then gives absolute convergence for Re⁡s>Re⁡s0+1\operatorname{Re}s>\operatorname{Re}s_0+1.

Local uniform convergence

Within its open half-plane of convergence, a Dirichlet series converges uniformly on compact subsets. This form of uniform convergence implies that its sum is an analytic function there.

Uniform convergence on compact subsets should not be confused with uniform convergence on an entire unbounded half-plane. The latter is a stronger property and is studied through a separate abscissa of uniform convergence.

Principal examples

Riemann zeta function

The defining series

ζ(s)=∑n=1∞n−s\zeta(s)=\sum_{n=1}^{\infty}n^{-s}

converges absolutely for Re⁡s>1\operatorname{Re}s>1, and

σc=σa=1.\sigma_c=\sigma_a=1.

In this half-plane it has the Euler product

ζ(s)=∏p(1−p−s)−1,\zeta(s)=\prod_{p}(1-p^{-s})^{-1},

where pp ranges over all prime numbers.

The product follows from expanding each factor as a geometric series and applying the unique factorization of positive integers. It provides the basic connection between Dirichlet series and prime numbers.

Alternating zeta series

The Dirichlet eta function is given by

η(s)=∑n=1∞(−1)n−1ns.\eta(s)=\sum_{n=1}^{\infty}\frac{(-1)^{n-1}}{n^s}.

It converges for Re⁡s>0\operatorname{Re}s>0, but converges absolutely only for Re⁡s>1\operatorname{Re}s>1. Hence

σc=0,σa=1.\sigma_c=0,\qquad \sigma_a=1.

This realizes the largest possible gap between the two finite abscissas.

For Re⁡s>1\operatorname{Re}s>1, separating the even and odd terms gives

η(s)=(1−21−s)ζ(s).\eta(s)=(1-2^{1-s})\zeta(s).

The identity also plays a role in extending the zeta function beyond its original half-plane of convergence.

Dirichlet LL-series

A Dirichlet character χ\chi modulo a positive integer qq determines the Dirichlet LL-series

L(s,χ)=∑n=1∞χ(n)ns.L(s,\chi)=\sum_{n=1}^{\infty}\frac{\chi(n)}{n^s}.

For Re⁡s>1\operatorname{Re}s>1,

L(s,χ)=∏p(1−χ(p)p−s)−1.L(s,\chi)=\prod_p(1-\chi(p)p^{-s})^{-1}.

A nonprincipal character has bounded partial sums, so its series converges throughout Re⁡s>0\operatorname{Re}s>0. These functions are fundamental to the study of primes in arithmetic progressions.

Coefficient algebra

Dirichlet convolution

Suppose

F(s)=∑n≥1ann−s,G(s)=∑n≥1bnn−s.F(s)=\sum_{n\geq1}a_n n^{-s}, \qquad G(s)=\sum_{n\geq1}b_n n^{-s}.

In a half-plane where both series converge absolutely,

F(s)G(s)=∑n≥1cnn−s,F(s)G(s)=\sum_{n\geq1}c_n n^{-s},

where

cn=∑d∣nadbn/d.c_n=\sum_{d\mid n}a_d b_{n/d}.

This coefficient operation is called Dirichlet convolution.

For example,

ζ(s)2=∑n=1∞d(n)ns,Re⁡s>1,\zeta(s)^2=\sum_{n=1}^{\infty}\frac{d(n)}{n^s}, \qquad \operatorname{Re}s>1,

where d(n)d(n) is the number of positive divisors of nn.

If a1≠0a_1\neq0, the sequence (an)(a_n) has a unique inverse under Dirichlet convolution. This is an algebraic statement; representing the reciprocal function 1/F(s)1/F(s) by a convergent Dirichlet series requires an appropriate convergence region.

Multiplicative coefficients and Euler products

An arithmetic function aa is multiplicative if a1=1a_1=1 and

amn=amanwhenever gcd⁡(m,n)=1.a_{mn}=a_m a_n \quad\text{whenever }\gcd(m,n)=1.

When its Dirichlet series converges absolutely, it factors as

∑n=1∞anns=∏p(∑k=0∞apkpks).\sum_{n=1}^{\infty}\frac{a_n}{n^s} = \prod_p \left(\sum_{k=0}^{\infty}\frac{a_{p^k}}{p^{ks}}\right).

If aa is completely multiplicative, so that the multiplication rule holds without the coprimality condition, this simplifies to

∏p(1−app−s)−1.\prod_p(1-a_p p^{-s})^{-1}.

Not every Dirichlet series has an Euler product. The factorization reflects a specific multiplicative structure in its coefficients.

Differentiation and uniqueness

Inside the half-plane of absolute convergence, termwise differentiation gives

F′(s)=−∑n=1∞anlog⁡nns.F'(s)=-\sum_{n=1}^{\infty}\frac{a_n\log n}{n^s}.

Higher derivatives introduce higher powers of −log⁡n-\log n.

The coefficients are uniquely determined by the represented function. For a series convergent in some right half-plane, they can be recovered successively along the real axis:

a1=lim⁡σ→+∞F(σ),a_1=\lim_{\sigma\to+\infty}F(\sigma),

and, for m≥2m\geq2,

am=lim⁡σ→+∞mσ(F(σ)−∑n<mann−σ).a_m=\lim_{\sigma\to+\infty} m^\sigma \left( F(\sigma)-\sum_{n<m}a_n n^{-\sigma} \right).

Thus two Dirichlet series agreeing on a right half-plane have identical coefficients.

Summatory functions and integral representations

Define the coefficient summatory function

A(x)=∑n≤xan.A(x)=\sum_{n\leq x}a_n.

Summation by parts gives, for x≥1x\geq1,

∑n≤xanns=A(x)x−s+s∫1xA(u)u−s−1 du.\sum_{n\leq x}\frac{a_n}{n^s} = A(x)x^{-s} +s\int_1^x A(u)u^{-s-1}\,du.

If the boundary term tends to zero and the improper integral converges, this becomes

F(s)=s∫1∞A(x)x−s−1 dx.F(s)=s\int_1^\infty A(x)x^{-s-1}\,dx.

For example, if

A(x)=O(xθ),A(x)=O(x^\theta),

then the representation proves convergence for Re⁡s>θ\operatorname{Re}s>\theta. Here OO is big-O notation. Cancellation in A(x)A(x), rather than the sizes of the individual coefficients alone, can therefore enlarge the half-plane of convergence.

After the substitution x=eux=e^u, the same formula reads

F(s)=s∫0∞A(eu)e−su du,F(s)=s\int_0^\infty A(e^u)e^{-su}\,du,

connecting the series to the Laplace transform. The preceding integral also has the form of a Mellin transform.

In the reverse direction, Perron-type inversion formulas express sums of coefficients through contour integrals involving F(s)xs/sF(s)x^s/s. Their use requires suitable convergence or truncation conditions and care at integer values of xx, where endpoint conventions affect the contribution of axa_x.

Analytic continuation and applications

The region where a Dirichlet series converges need not be the largest region where its sum can be defined analytically. Analytic continuation extends the represented function, not the convergence of the original series.

For example, the defining series for ζ(s)\zeta(s) converges only when Re⁡s>1\operatorname{Re}s>1, but the zeta function extends meromorphically to the whole complex plane, with a simple pole at s=1s=1. Values obtained elsewhere by continuation are not ordinary sums of the divergent defining series.

Dirichlet series allow analytic properties—poles, zeros, growth estimates, and functional equations—to be translated into information about arithmetic coefficients. A central example is

−ζ′(s)ζ(s)=∑n=1∞Λ(n)ns,Re⁡s>1,-\frac{\zeta'(s)}{\zeta(s)} = \sum_{n=1}^{\infty}\frac{\Lambda(n)}{n^s}, \qquad \operatorname{Re}s>1,

where the von Mangoldt function satisfies Λ(pk)=log⁡p\Lambda(p^k)=\log p for prime powers and is zero otherwise. This identity connects the zeros and poles of the zeta function with prime-counting questions and underlies analytic approaches to the prime number theorem.

Analytic continuation, Euler products, and functional equations are additional structures, not automatic consequences of being a Dirichlet series. An arbitrary coefficient sequence may yield a function with no continuation across part or all of its convergence boundary. Even when continuation exists, extracting coefficient asymptotics generally requires growth bounds and control of singularities beyond the continuation itself.

Generalized Dirichlet series

For a generalized series

F(s)=∑n=1∞ane−λns,F(s)=\sum_{n=1}^{\infty}a_n e^{-\lambda_n s},

the distribution of the frequencies λn\lambda_n affects convergence and coefficient operations. The bound σa−σc≤1\sigma_a-\sigma_c\leq1 for ordinary Dirichlet series does not apply without further assumptions on these frequencies.

Choosing λn=n\lambda_n=n gives

F(s)=∑n=1∞an(e−s)n,F(s)=\sum_{n=1}^{\infty}a_n(e^{-s})^n,

which becomes a power series under z=e−sz=e^{-s}. Choosing λn=log⁡n\lambda_n=\log n instead gives the ordinary Dirichlet series and its distinctive arithmetic structure. Along a vertical line s=σ+its=\sigma+it, the factors e−itλne^{-it\lambda_n} also relate generalized Dirichlet series to exponential sums and harmonic analysis.